Cellares logo

Catching deviations before batches fail: we added foresight to Cellares' production.

Cellares is an Integrated Development and Manufacturing Organization (IDMO) that develops and operates automated technologies for cell therapy manufacturing, helping therapy developers bring life-saving treatments to patients at scale.

With Cellares' product team, Wildstream implemented predictive monitoring on production telemetry and QA signals, giving operators early warnings of process deviations in robotic manufacturing.

Cellares
  • AI Engineering

Warnings before failure

Operators are warned of process deviations early, while there is still time to intervene before a batch fails

Two signal streams

Production telemetry and quality-assurance signals are monitored together to catch deviations that one source alone could miss

Patient-specific stakes

Each batch is made for a single patient, so saving a batch protects a treatment that cannot simply be replaced

AI Engineering

Predictive Process Monitoring

In automated cell-therapy manufacturing, small environmental deviations can make patient-specific batches fail quality checks. Robotic cell therapy manufacturing requires strict environmental control, and even a small deviation can push a batch out of specification. A failed batch can delay treatment for the patient it was made for.

Warnings before failure

Operators are warned of process deviations early, while there is still time to intervene before a batch fails.

Two signal streams

Production telemetry and quality-assurance signals are monitored together to catch deviations that one source alone could miss.

Patient-specific stakes

Each batch is made for a single patient, so saving a batch protects a treatment that cannot simply be replaced.

What we did

Statistical anomaly detection and machine learning on production telemetry and QA signals warn operators early, so they can intervene before a batch fails.

Sponsored by Cellares' product management, Wildstream implemented predictive monitoring for robotic cell therapy manufacturing. Statistical anomaly detection and machine learning run on production telemetry and QA signals, and operators receive an early warning when a process starts to deviate, so they can adjust it before the batch reaches its quality checks.

How it works
  1. Production telemetry is collected from the robotic manufacturing process
  2. QA signals are combined with the telemetry
  3. Statistical anomaly detection compares readings with normal conditions
  4. Machine learning predicts deviations that could make a batch fail
  5. Operators receive an early warning and intervene
Key capabilities

Telemetry monitoring

Data from robotic production is tracked continuously, giving a live view of environmental conditions during each run.

Statistical anomaly detection

Statistical models flag readings that drift away from normal conditions, even while the change is still small.

Predictive deviation alerts

Machine learning anticipates process deviations from telemetry and QA signals and gives operators a timely warning to act.

Technology
  • Statistical anomaly detection
  • Machine learning
Impact

Early warnings from production data that help operators save patient-specific cell therapy batches

Fewer failed batches

Intervening early helps keep patient-specific batches within quality limits instead of losing them at final quality checks.

Confident operators

Operators know where to look and when to act, rather than discovering a problem only after a batch has failed.

More reliable production

Fewer late surprises make cell therapy manufacturing runs more predictable for both the production team and the patients.

Bring us a challenge like this

Tell us where your operations still depend on manual work, and we will tell you what is worth building first.